September 2002 - Darwin-kernel

Re: Darwin & SMT?
by Jim Magee 10 Sep '02

10 Sep '02

On Tuesday, September 10, 2002, at 01:16 PM, Bernie Zenis wrote: Is anyone here familiar with Simultaneous Multithreading (http://www.cs.washington.edu/research/smt/)? Does anyone know if Darwin would lend itself well to SMT? For example, if a SMT PowerPC chip came out, would it be easy to port Darwin to it? I realize that might depend on what kind of SMT architecture was implemented. I have always felt that Darwin works best on an SMP machine. There are enough system threads, and multi-threaded applications to want two CPUs in many "peak" CPU utilization situations (my Mac OS X system typically has ~200 threads). I think the real world anecdotal evidence bears this out as well (most people with dual-ies are quite happy with the responsiveness of their machines, even when compared to single processor machines of quite higher clock rates). All of this is one of the reasons Apple ships all their PowerMacs as dual processor machines right now. The CPI numbers under most of those situations indicate that an SMT (others call it Hyper-threading) processor might do reasonably well compared to an SMP machine. But as you say, it all depends on the particular SMT design and other architectural factors on any given processor. --Jim _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Darwin & SMT?
by Bernie Zenis 10 Sep '02

10 Sep '02

Hi, Is anyone here familiar with Simultaneous Multithreading (http://www.cs.washington.edu/research/smt/)? Does anyone know if Darwin would lend itself well to SMT? For example, if a SMT PowerPC chip came out, would it be easy to port Darwin to it? I realize that might depend on what kind of SMT architecture was implemented. Thanks, Bernie _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Can Darwin Live Without Disks?
by Bernie Zenis 10 Sep '02

10 Sep '02

Thank you, everyone. I think all of my questions have been answered even if I didn't make the best sense asking them. :-) I appreciate the help. -Bernie _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Can Darwin Live Without Disks?
by Emiel Kollof 10 Sep '02

10 Sep '02

On Monday 09 September 2002 19:36, Bernie Zenis wrote: [sxplanation snipped] > So, it isn't quite there yet; but, it could be? Does anybody out there > think that Darwin will eventually have the capabilities to work from a > diskless (or from a non writable disk) machine? I believe Mac OS X (and therefore also Darwin) can be netbooted. Apple sells just this thing with their Xserve machines, and it works like I described. It uses some odd DHCP configuration, and some folks already reverse engineered it. Please check out this URL: http://mike.passwall.com/macnc/ > I know hardware is a lot more robust these days; however, I think that > it would be a great selling point from the Mac to have an OS that runs > without disks. "Without disks" can mean a user/sys admin decides not to > use a disk, the disk crashes, power is out and the disk(s) are "shut > down" with the system still going, or whatever you can think of. Maybe > it would only really be a selling point for Mac servers and computer > geeks (like me :-). I have had much luck with netbooting NetBSD on my iMac. OpenFirmware knows DHCP and tftp, so that Just Worked(tm). It's not impossible, in fact, it's very possible, maybe even doable to netboot Darwin/OS X from another *nix box. The only thing it needs for it being more useable is better documentation. It's not rocket science :) > > Hope that cleared stuff up a bit. > > Yes, thank you. No problem :) Cheers, Emiel -- To invent, you need a good imagination and a pile of junk. -- Thomas Edison _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Can Darwin Live Without Disks?
by Bernie Zenis 09 Sep '02

09 Sep '02

On Sunday, September 8, 2002, at 08:02 PM, Emiel Kollof wrote: * Bernie Zenis (bzenis(a)mac.com) wrote: I've heard of netbooting but I don't know any details. Is nothing loaded from a local disk? That is, does firmware make a network connection and download the needed files? Does a netbooted machine still need a local writable disk? Can VM work over a network? Does Darwin need to have swap file(s)? If RAM is large enough, can everything happen over the network? Netbooting a UNIX usually depends on a BOOTP/DHCP/RARP to get an ip number (which protocol is used depends on the hardware). Open Firmware on newer (New world) macs is pretty content with DHCP. Then with tftp a boot image is downloaded. This boot image contains something that understands some remote file system like perhaps Appletalk or NFS. NFS is more common. The boot image in turn fetched the kernel from that network filesystem and boots it. Usually there are these approches (assuming NFS): 1. The kernel knows NFS and can use a NFS root right away, and boots on. 2. A ramdisk is set up with a barebones root and boots on. Swapping over NFS is quite common, but not recommended. Disks are slow, but networks are usually slower. The way Mac OS X/Darwin works is probably along those lines. So yes, in theory it could function without a disk, given enough tinkering. So, it isn't quite there yet; but, it could be? Does anybody out there think that Darwin will eventually have the capabilities to work from a diskless (or from a non writable disk) machine? I know hardware is a lot more robust these days; however, I think that it would be a great selling point from the Mac to have an OS that runs without disks. "Without disks" can mean a user/sys admin decides not to use a disk, the disk crashes, power is out and the disk(s) are "shut down" with the system still going, or whatever you can think of. Maybe it would only really be a selling point for Mac servers and computer geeks (like me :-). Hope that cleared stuff up a bit. Yes, thank you. -- Dave Barry, "The Taming of the Screw" LOL -Bernie _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Can Darwin Live Without Disks?
by Ragnar Sundblad 09 Sep '02

09 Sep '02

--On Sat, 7 Sep 2002 20:08:05 -0400 Bernie Zenis <bzenis(a)mac.com> wrote: What do you mean by disk? The OS has to be read from something and it doesn't have to be a "disk" in the traditional sense of the word. Hard disk, CD, DVD, ZIP, etc. Not that a CD is write protected, and you can boot from it. I've heard of netbooting but I don't know any details. The outlines for this was described in another email, that is more or less how it works with darwin too (with darwin the kernel and the drivers are tftp-ed too, by the boot loader). Is nothing loaded from a local disk? Not necessarily, only if you set it up to do that. That is, does firmware make a network connection and download the needed files? Yes, OpenFirmware knows how to do this. Does a netbooted machine still need a local writable disk? Not necessarily, but see the swap issues below. Can VM work over a network? Probably, not sure, but that could be slow. (With a good server and gig ethernet you maybe could get pretty good performance though.) Does Darwin need to have swap file(s)? No, but many unix programs aren't really good on handling the case when there is no more memory to allocate, so you wouldn't want to do this for a general purpose user machine. (Mac OS 9 apps are usually better on handling that since it was something that happened every now and then.) This is not a problem with the kernel as much as with the applications that you may want to run, and probably also the libraries that they use. If RAM is large enough, can everything happen over the network? Certainly, but see above. For a general purpose machine, you will eat up all your RAM sooner or later, and things will probably start to fail. If the "disk" that contains the OS image and more importantly the VM swap space goes away the system will most likely fail soon afterwards. That is the answer I was expecting. I just want to know how robust or fail-safe Darwin is. I haven't tried it, but if you boot of a firewire disk and disconnects it, I wouldn't be surprised if you got something telling you to reconnect the disk so that the machine can continue doing things. If it has been allowed to swap things out it will probably need some of those pages again sometimes later. I think even parts of the kernel can be swapped out now adays. I'm mostly seeking knowledge. If Darwin can startup and run without a disk, then the whole Mac OS is closer to being able to do the same, Note that it can boot from a CD. The Mac OS X installer is running under Mac OS X. Check out the new Mac OS X Server Netboot stuff, I am not sure if they are swapping localy or not, you can probably set it up either way. then it might be useful to write Mac software that is "diskless friendly". I don't know what you mean by that, but it is never correct to assume that all disk accesses will work or that you can write everywhere you are used to, there can be hardware or permission problems for example. I guess a better, more general question might be, "Can Darwin startup and run without a writable disk, local or remote?" Yes, CD or NFS for example. Does any of this make sense to you or just to me? I don't really understand what you want to know. There is a spectrum of boot options, but they are not all recommendable for all kinds of use. A diskess Mac OS X lab room computer would probably be pretty useless if it couldn't swap, a darwin based ATM machine maybe wouldn't. Note that you need some kind of filesystem somewhere to get to the software to run. Otherwise you will only have a kernel with some drivers which isn't all that interresting (for most of us at least, maybe on this list there are people that actually have use for that :-). /ragge _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Physical Memory, DMA, and I/O Mappings
by Josh de Cesare 09 Sep '02

09 Sep '02

We are in the process of making changes to the way the kernel handles I/O mappings and physical memory addresses. The changes will be released in a future update to Mac OS X 10.2. We have found that the changes effect various drivers and will require those drivers to change. Below is a description of each of the major changes and suggestions about how drivers should be adapted. We do not expect any performance issue or backward compatibility problems with drivers once they have been adapted. I/O Mappings: Currently, the kernel maps PCI I/O & memory spaces such that the kernel virtual address equals the physical address (V=R). Some drivers access their registers by obtaining the physical address of the registers then using the address as a normal virtual address. This has never been correct, but has worked because of the V=R I/O mappings. These extra mappings will no longer be available. Drivers should create an explicit mapping for their memory and I/O spaces. This can be done easily with methods like provider->mapDeviceMemoryWithIndex. This will look up the range in the provider and map it for the driver. The driver can then get the virtual address. Some families have similar methods: IOPCIFamily provides mapDeviceMemoryWithRegister so that the order of the spaces need not be known, just which base address register the hardware uses. pmap_extract and kvtophys: These functions currently find the physical address for a page given a virtual address and a page. Drivers often use this physical address to setup DMA programs. Normally a driver uses physical page addresses for two purposes: a small number of pages to hold the DMA programs, and then a large number of pages that contain the data for the I/O. pmap_extract and kvtophys will no longer be available for use in drivers. Drivers should use IOMemoryDescriptors to get physical page addresses. For pages used to hold DMA programs, IOBufferMemoryDescriptor will allocate and map memory for the driver. getPhysicalSegment, or getPhysicalAddress will then return the physical address to be given to DMA hardware. The driver must call ->prepare() on the IOBufferMemoryDescriptor before it is used. Call ->complete() and then ->release() when the memory is no longer needed. Assuming the IOBufferMemoryDescriptor has been allocated into the kernel task, the memory can be accessed normally by getting its address with ->getBytesNoCopy(). If the driver does not map the memory, it can still be accessed with ->readBytes(...) and ->writeBytes(...). The I/O memory for most drivers will already be in a IOMemoryDescriptor. If the data is not provided in an IOMemoryDescriptor, the driver should wrap it with one; IOMemoryDescriptor::withAddress(...). Use a loop across getPhysicalSegment to determine the physical address of each page. Be aware the getPhysicalSegment will also return a length, which can be of any size and can be larger or smaller than one page. Drivers may also use IOMemoryCursor to simplify the iteration process. -- Josh de Cesare Mac OS X - Core OS Apple Computer, Inc. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: X.2 mach abort bug
by Jim Magee 09 Sep '02

09 Sep '02

On Monday, September 9, 2002, at 12:23 AM, devdrvr.net Multi-Platform SW wrote: #0 0x9001bc0c in kill () #1 0x9005d5cc in abort () #2 0x90006ac4 in _lu_fork_child () #3 0x9000d410 in _lu_running () #4 0x9008a560 in getserv () #5 0x001ae3f0 in serv_name_lookup (port=49229, proto=PT_NONE) at resolv.c:236 Using libpcap-0.7.1's pcap_read() in my code I sometimes get this abort later up the stack. Now I can't get it to happen at all (after trying sans windowing system and going back to Aqua) and I forgot the exact mach port abort message that got spit out at me. resolv.c is not mine. It's not really a Mach error. The dyld subsystem failed [to do on-the-fly symbol resolution] in a place where it never should. So, aborting was its only option. The question is: Why? It could be a reentrancy problem. You aren't calling this from a signal handler, are you? Anything that does these kinds of library calls are not safe from a signal handler. But without information about what stack frames 6->N are, there is no good way to tell from here. --Jim _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

X.2 mach abort bug
by devdrvr.net Multi-Platform SW 09 Sep '02

09 Sep '02

#0 0x9001bc0c in kill () #1 0x9005d5cc in abort () #2 0x90006ac4 in _lu_fork_child () #3 0x9000d410 in _lu_running () #4 0x9008a560 in getserv () #5 0x001ae3f0 in serv_name_lookup (port=49229, proto=PT_NONE) at resolv.c:236 Using libpcap-0.7.1's pcap_read() in my code I sometimes get this abort later up the stack. Now I can't get it to happen at all (after trying sans windowing system and going back to Aqua) and I forgot the exact mach port abort message that got spit out at me. resolv.c is not mine. At least the symbolics file with gdb thinks it knows the line #. If you've got access to this file for Darwin I'd be interested to know what is going on. May be a bug. Weird. It was happening every time through when I was looking at my Airport card port on my TiBook. --Perry http://www.devdrvr.net/PerryCard.html _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Can Darwin Live Without Disks?
by Emiel Kollof 09 Sep '02

09 Sep '02

* Bernie Zenis (bzenis(a)mac.com) wrote: > > I've heard of netbooting but I don't know any details. Is nothing loaded > from a local disk? That is, does firmware make a network connection and > download the needed files? Does a netbooted machine still need a local > writable disk? Can VM work over a network? Does Darwin need to have swap > file(s)? If RAM is large enough, can everything happen over the network? Netbooting a UNIX usually depends on a BOOTP/DHCP/RARP to get an ip number (which protocol is used depends on the hardware). Open Firmware on newer (New world) macs is pretty content with DHCP. Then with tftp a boot image is downloaded. This boot image contains something that understands some remote file system like perhaps Appletalk or NFS. NFS is more common. The boot image in turn fetched the kernel from that network filesystem and boots it. Usually there are these approches (assuming NFS): 1. The kernel knows NFS and can use a NFS root right away, and boots on. 2. A ramdisk is set up with a barebones root and boots on. Swapping over NFS is quite common, but not recommended. Disks are slow, but networks are usually slower. The way Mac OS X/Darwin works is probably along those lines. So yes, in theory it could function without a disk, given enough tinkering. Hope that cleared stuff up a bit. Cheers, Emiel -- Electricity is actually made up of extremely tiny particles, called electrons, that you cannot see with the naked eye unless you have been drinking. Electrons travel at the speed of light, which in most American homes is 110 volts per hour. This is very fast. In the time it has taken you to read this sentence so far, an electron could have traveled all the way from San Francisco to Hackensack, New Jersey, although God alone knows why it would want to. The five main kinds of electricity are alternating current, direct current, lightning, static, and European. Most American homes have alternating current, which means that the electricity goes in one direction for a while, then goes in the other direction. This prevents harmful electron buildup in the wires. -- Dave Barry, "The Taming of the Screw" _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0